A reusable methodology for converting vague executive briefs into structured, executable program plans — using AI as a thinking partner, not a replacement for TPM judgment.
Executive briefs are often light on decisions and heavy on intent. Words like "TBD," "maybe," and "I think it's resolved" appear where scope, budget, and legal clearance should be. Teams that treat these briefs as execution-ready start building on unvalidated foundations.
The TPM's job is to make ambiguity executable — before engineering starts, before money is spent, before assumptions become technical debt.
| File | Description |
|---|---|
ai_initiative_brief.md |
A fictional CPO brief — realistic but made up |
AI_Initiative_TPM_Analysis.md |
Full structured program analysis from that brief |
CLAUDE.md |
The methodology: reasoning log, judgment calls, 7-step reuse pattern |
Note: The brief used in this example is fictional — created to demonstrate the methodology. The methodology is real and generalizable to any executive vision brief.
Documented in full in CLAUDE.md. The short version:
- Characterize the document type — vision brief, spec, strategy memo each require different responses
- Run three filters — stated facts / assumptions / unresolved decisions
- Flag "I think" statements on compliance, legal, or dependency items — these are almost always unverified
- Decompose workstreams by separation of concerns, not by org chart
- Anchor phasing to hard constraints first, then fill in the rest
- Split metrics into leading and lagging — leading indicators save programs, lagging indicators evaluate them
- End with concrete immediate actions — no more than 5, owned and time-bound
Most TPMs share the analysis. This repo shares the meta-layer: a CLAUDE.md
that documents how the analysis was made — which judgment calls were made,
why, and how to replicate the thinking on a different brief.
This is the artifact worth studying. The analysis is one output. The reasoning log is a reusable thinking tool.
- Take any vague executive brief
- Read
CLAUDE.md— understand the 7-step pattern and the judgment call table - Apply the three filters to your brief (stated / assumed / unresolved)
- Use the workstream decomposition and phasing approach as a starting structure
- Adapt the success metrics framework to your domain
AI accelerates steps 3-5 significantly when given the right context. The judgment in steps 1-2 remains human.
This methodology maps directly to core TPM competencies:
- Ambiguity resolution — converting "let's make it happen" into a program charter
- Risk identification — surfacing blockers before they become crises
- Stakeholder alignment — structured open questions force decisions that would otherwise be deferred
- Cross-functional coordination — workstream decomposition by separation of concerns, not org chart
The pattern scales: platform migrations, vendor evaluations, org restructuring, regulatory programs — any domain where a senior leader has a vision and a team needs to execute it.
Built by Joanna — TPM specialising in AI/ML, agentic workflows, and AI Safety operations. Exploring what it means to make AI ambiguity executable.
AgentRed-Light — a related project: guardrail test suite for AI agents.